Train a reference torch model through luz

The caller supplies case-disjoint training and validation Datasets. When subject metadata is attached with withCaseData(), both datasets must carry it and subjects must also be disjoint. Legacy unannotated inputs retain case-only checking; missing metadata does not establish subject separation. Validation must reuse training normalization statistics. CPU execution, no worker processes, and deterministic caller-state restoration are enforced in this implementation.

Usage

trainModel(
  train,
  valid = NULL,
  model = c("eegnet", "cnn1d", "tcn", "lstm"),
  epochs = 20L,
  batch_size = 32L,
  learning_rate = 0.001,
  weight_decay = 0,
  callbacks = NULL,
  seed = 1L,
  device = "cpu",
  num_workers = 0L,
  verbose = FALSE,
  ...
)

Arguments

train, valid: Targeted physio_torch_dataset objects.
model: Exact architecture name or a physio_torch_module.
epochs, batch_size: Positive whole-number training settings.
learning_rate: Positive Adam learning rate.
weight_decay: Non-negative Adam weight decay.
callbacks: NULL or a list of luz callback instances.
seed: Whole-number CPU seed.
device: Currently exactly "cpu".
num_workers: Currently exactly zero.
verbose: Whether luz prints progress.
...: Architecture settings when model is a name.

Value

A physio_dl_model.

Examples

arr <- array(stats::rnorm(50 * 2 * 6), dim = c(50, 2, 6))
pe <- PhysioExperiment::PhysioExperiment(
  assays = list(raw = arr),
  colData = S4Vectors::DataFrame(label = c("C3", "C4")),
  samplingRate = 100
)
# Training needs the optional torch and luz backends.
if (requireNamespace("torch", quietly = TRUE) &&
    requireNamespace("luz", quietly = TRUE)) {
  train <- peDataset(
    pe, targets = rep(c("a", "b"), 3),
    task = "classification", class_levels = c("a", "b")
  )
  fit <- trainModel(train, model = "cnn1d", epochs = 1L)
}
#> Error: the R package `torch` is installed but LibTorch is unavailable; install it explicitly with `torch::install_torch()`